Crystal Palace vs Arsenal

Start with the averages. Crystal Palace's home matches average 8.7 corners per game, Arsenal's away matches average 9.3, and both sit below the Premier League baseline of 10.5. The model's predicted total of 9.88 reflects that suppressed picture, but the variance in either team's recent run complicates a clean read.

Arsenal's last five away totals tell a compressed story: 7, 7, 13, 7, 7. Four of five at 7. One outlier at 13 pulls the average upward and raises a question the model has to absorb: is the 13 a structural signal or noise? Crystal Palace's last five at home show a wider spread, 7, 15, 10, 6, 11, and that kind of dispersion at Selhurst is what nudges the pick into medium territory rather than the high confidence tier, where the model's own output sits.

The gap between model confidence and pick confidence matters. High model confidence means the predicted total is stable across the ensemble's internal checks; medium pick confidence means the situational context introduces enough uncertainty that we don't weight this identically. Selhurst Park in a late-season London derby carries crowd and tactical variables that rolling averages can't fully price. We're not dismissing the signal.

Stake discipline applies, full stop. Medium confidence carries a defined weight for a reason; it's not a signal to ignore or to overload. In the Premier League, medium confidence picks have settled 134 times at a 45.5% win rate.